Closed pmarks closed 4 years ago
@pmarks : I thought I left a comment on this thread, but now can't find it. Anyway, enclone passes tests after using this and is faster, although I don't know how much of the speedup was due to the avx2 and how much due to the ndiffs change.
apologies, this got entangled with a cargo fmt. Interesting method is here: https://github.com/10XGenomics/rust-debruijn/compare/pmarks/faster-construct?expand=1#diff-d7e4ec435e97e4f6ab2d5d4b07bfe07cR149
Constructing a DNA string from a &[u8] of ACGTs is now >10x faster. Could possibly be extended to other construction methods if that's useful. A regression of this branch against enclone would be great, although there are a bunch of unit tests.